Policing of realistic VBR video traffic in an ATM network

Abstract

AbstractATM has been accepted by CCITT as the transport mechanism for the future BISDN and will also be widely used in future customer premises networks. Networks based on the ATM principle are expected to provide a very flexible communications infrastructure allowing customers to make effective use of a wide variety of offered services. To provide this flexibility with an acceptable quality of service while operating the network in an economic way, elaborate traffic management functions will be necessary to control the traffic flows within the network. This paper will study one of these functions—the so‐called ‘usage parameter control’ or ‘policing’ function—in some detail to illustrate some of the problems that arise and point out possible solutions. The mechanisms chosen to implement the policing function will be the ‘leaky bucket’ mechanism, the ‘jumping window’ mechanism and the ‘moving window’ mechanism. The input streams used to assess the mechanisms represent different types of video communication—videophone, video conference and entertainment video—coded according to different variable bit‐rate (VBR) algorithms. In contrast to most of the previous studies, where artificial, statistical traffic sources have been used, the sources used in this paper are directly based on measured ‘real‐life’ video data. This ensures that all the statistical properties of the actual traffic stream are preserved and allows identification of the different factors that influence the dimensioning and the performance of the policing mechanism. The results of this study show that the uncertainty about the key parameters at call set‐up and the considerable impact of single scenes make the proper dimensioning of policing mechanisms difficult. Furthermore, it seems not to be practical to use the long term mean bit‐rate as the key traffic control parameter for these sources. Results indicating that the long‐term cell loss ratio is not a sufficient measure for the quality of service are also presented. A comparison of the mechanisms shows that from a performance perspective, the ‘leaky bucket’ mechanism is superior to the two window mechanisms. This work is relevant to evolving standards for both BISDN traffic management and variable bit‐rate video coding.